SparseWake/sparsewake
SparseWake SparseWake is a synthetic benchmark for sparse temporal hydrodynamic sensing. ICLR 2027 release The expanded release adds controlled multi-source mixtures and common-prior nearest-source tasks, with complete core data banks, reference checkpoints, a small review supplement, and reproduction code with a frozen wake-library input. Download release iclr2027-v1.0rc2 The version page lists the three archives, exact sizes, checksums, extraction instructions… See the full description on the dataset page: https://huggingface.co/datasets/SparseWake/sparsewake.
0275
1from __future__ import annotations2 3import math4 5import numpy as np6from sklearn.model_selection import train_test_split7 8 9def pose_holdout_split(10 pose_id: np.ndarray,11 seed: int = 1,12 test_fraction: float = 0.2,13 val_fraction: float = 0.15,14) -> tuple[np.ndarray, np.ndarray, np.ndarray]:15 rng = np.random.default_rng(seed)16 poses = np.unique(pose_id)17 n_test = max(1, int(math.ceil(test_fraction * len(poses))))18 test_poses = rng.choice(poses, size=n_test, replace=False)19 sample_idx = np.arange(len(pose_id), dtype=np.int64)20 test = sample_idx[np.isin(pose_id, test_poses)]21 remaining = sample_idx[~np.isin(pose_id, test_poses)]22 train, val = train_test_split(remaining, test_size=val_fraction, random_state=seed)23 return train, val, test24